From DevOps to MLOps: Overview and Application to Electricity Market Forecasting
Round 1
Reviewer 1 Report
The explanation of the latest DevOps and MLOps and the introduction of the MLOps framework as a case study looks to be a good study.
My concern is that it's hard to find scientific research contributions. Sections 2 and 3 consist mostly of introductions of existing theories and techniques, and it might be difficult to find technical challenges and scientific contributions in Section 4, which is the main contribution of this paper.
Would it be possible to make the contents of Section 2 and 3 more compact?
In Conclusion Section, it is mentioned that generalizing pipeline structure is difficult, however there is no clear explanation as to why this is difficult. Moreover, it is hard to find an explanation of how Section 4 solved this difficulty. It would be a strong paper if you explain above.
Author Response
Dear Reviewer,
Thanks for your review. Our response is in the attached word document.
Best Regards,
Authors
Author Response File: Author Response.docx
Reviewer 2 Report
The paper is well written in English, but the DevOps subject is not new. Thus, the entire section 3.1 can be eliminated - there is no new information regarding DevOps practices.
The novelty of the paper is not highlighted.
Figure 3 - Graphic representation of the feature branching strategy is not new, is adapted from 109, 110. This should be mention on the Figure 3 legend (if the section 3.1 remains).
Figure 4 - Graphic representation of trunk-based versioning is not new, is adapted from 111, 112. This should be mention on the Figure 4 legend (if the section 3.1 remains).
The sentence 'A common market structure in Europe and elsewhere is the hourly day-ahead market' (line 640) has no references. This is all the more surprising, as the work abounds with references.
Reproducibility of the result is an important technical issues of the ML systems - line 445. But the paper did not explain how reproducibility issue is solved using the proposed MLOps. To be more specific, every time when a neural network is retrained, other results can be obtained. What are the proposed steps to solve this issue?
Regarding the references, nine pages of them seem a bit exaggerated.
Author Response
Dear Reviewer,
Thanks for your review. Our response is in the attached word document.
Best Regards,
Authors
Author Response File: Author Response.docx
Reviewer 3 Report
In this paper, they proposed overview and application to electricity market forecasting. The paper is well written. But there are some drawbacks regarding the paper as follows:
a)The abstract should be shortened and rewritten. It is mostly focused on the reason of the study. The authors should highlight the results and advantages of mentioned methods.
b) Why did you choose the ML models (ANN, SVM, KNN, and a Decision Tree)? Please explain it in the paper.
c) What is the motivation of your paper?
d) Please extend the results by adding new experiments.
e) About the literature, each paper should clearly specify what is the proposed methodology, novelty, and results.
f)Please state limitation (weaknesses) of the proposed approach (if any) in the conclusion section and future research direction should be elaborated in more detail.
Author Response
Dear Reviewer,
Thanks for your review. Our response is in the attached word document.
Best Regards,
Authors
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
All the comments have been addressed in the revised version.
Reviewer 3 Report
The authors have made the modifications suggested.